3.3.1 Measurement of Stock Return
Stock market return is taken from market index in each South-East Asia Country. Market index can represent of overall activity of each country. Market
Return can be calculated by stock price in period t minus stock price index in period t-1 divided by stock price period t-1. Market return can be formulated
mathematically as Pisedtasalasai and Gunasekarage, 2008;
……………………...……………… 1 where;
: Market Return in period t : Market Price in period t
: Market Price in periodt-1
3.3.2 Measurement of De-trended Volume
The first variable of this research is return. According Choi et al 2012, trading volume has influence to return, and vice versa. To make regression model,
these two variables should be in the same form. If the return is using percentage form, trading volume should be in the percentage form too. Thus, the form of
trading volume will be formulated, as following Pisedtasalasai and Gunasekarage, 2008;
………………………………….…. 2 where;
DV
t
= De-trended Volume in period t Trading Volume
t
= Trading Volume in period t Trading Volume
t-1
= Trading Volume in period t-1
3.4 Data Analysis Method
3.4.1 Vector Auto-Regression VAR Analysis
Vector Auto-Regression VAR analysis was built by Sims 1980. VAR analysis is used to project the system variable time series data and analyze the
influence of dynamic disturbance contained in the equation. In this VAR model, it is not necessary to categorize which variable is endogen dependent or
exogenous independent. Sims 1980 assumes that all variable in VAR model is endogenous dependent. There is interdependent between variables. For instance,
variable A have influence to variable B. while in the same, variable B also have influence to variable A. It means that there is causality relationship between
variable A and variable B.
According Widarjono 2013, there are steps to run VAR analysis; 1 stationary test with the data, 2 Co-integration test, 3 determine maximum lag
and optimal lag which will be used, 4 Causality test, 5 estimation VAR, and 6 analyze result of Impulse Response and Variance Decomposition.
1. Stationary Test
This research adopts a test for a unit root test to ensure that variable is stationary, and to avoid spurious regression there is no relationship between
dependent variable and dependent variable. Stationary test can detect spurious regression. Stationary test can explain the behavior of the data too.
Therefore, it is important to stationary test for time series data.
The testing for a unit root is based on Augmented Dickey-Fuller 1979 ADF and Phillip-Perron 1988 PP. ADF and PP test are used with trend
and without trend. The ADF test formulated as follows Widarjono, 2005;
…………………………..... 3 Stationary data is based on statistical comparison from MacKinnon critical
value. If statistic value of ADF and PP test absolutely higher that Mackinnon critical value in level α 1, 5, and 10, so data is called stationary.
Analyzing using VAR model, data used should be stationary in the same level. If one of data is not stationary, data should be tested in the 1
st
difference or 2
nd
difference.
2. Determining optimal lag
The most important in VAR analysis is determining the lag length. The optimal lag is needed to catch the influence of each variable to other variable
in VAR model. There are five criteria can be used to determine the optimal lag; 1 Akaike Information Criterion AIC, 2 Schwartz Information
Criterion SIC, 3 Hannan-Quinn Information Criterion, 4 Likelihood Ratio LR, and 5 Final Prediction Error FPE.